FastInfra Solutions

AI video generation API.

Submit LTX-2.5 jobs with an OpenAI-compatible workflow — HTTP 202, poll for completion, MP4 output.

Why FastInfra

GET /videos/models — list video SKUs
POST /videos — async job submit (HTTP 202)
Poll GET /videos/{id} until completed
Configurable duration and resolution
Billed per second of output from credits
Documented in OpenAPI and llms.txt

How the video API works

Video generation is asynchronous because a clip takes longer to render than an HTTP request should stay open. You submit a job, receive a job ID immediately with HTTP 202, poll until it completes, then download the MP4.

GET https://api.fastinfra.ai/v1/videos/models — list the video model IDs you can use.
POST https://api.fastinfra.ai/v1/videos — submit a prompt with duration and resolution; the response is HTTP 202 with a job ID.
GET https://api.fastinfra.ai/v1/videos/{id} — poll until status is completed, then fetch the output URL.

curl https://api.fastinfra.ai/v1/videos \
  -H "Authorization: Bearer $FASTINFRA_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "lightricks/ltx-2.5", "prompt": "A drone shot over a misty pine forest at sunrise", "seconds": 5}'

Pricing

Video is billed per second of output from prepaid credits; output seconds are converted to token-equivalent usage, so video and LLM calls share one balance and one usage report. There is no subscription and no per-seat fee.

Prompting tips for better clips

Describe the camera (drone shot, close-up, slow pan) as well as the subject.
Name the lighting and time of day; it has a large effect on realism.
Keep one action per clip; generate several short clips and cut them together rather than one long, complex prompt.
Start at a low resolution to iterate on the prompt, then render the final take at full resolution.

Common use cases

Product and social media loops, B-roll for explainers, storyboards and pre-visualisation, and generated scenes inside apps — all driven from the same API key you use for LLMs.

Frequently asked questions

Is the video API OpenAI-compatible?

It follows a similar async job pattern. See the Docs video section for the exact request and poll contract.

How is video billed?

Output seconds convert to token-equivalent usage and debit prepaid credits.

Where is the machine-readable spec?

openapi.json and llms.txt on the marketing site document the endpoints.

How long does a video job take?

It depends on the model, duration and resolution; short clips usually complete in tens of seconds to a few minutes. Poll the job endpoint rather than holding the request open.

Can I use the OpenAI SDK for video?

Use plain HTTP or your HTTP client of choice for the async job endpoints; the same FastInfra API key and base URL work for video and chat.

Which resolutions are supported?

Supported sizes and durations are listed per model by GET /videos/models and in the Docs video section.

Video API docs